Introduction

This report aims to perform a full data visualisation and presents comprehensive analysis on the Use of Force (UOF) dataset from Dallas, Texas. The UOF dataset possesses information on the incidents where law enforcement officers used force against individuals during encounters, including details on the dates, location, officer and subject details, injuries sustained and other relevant information related to UOF incidents in 2016.

The purpose is to provide valuable insights with various trends and patterns in the form of tables, graphs and plots to gain a better understanding of the factors associated with the dataset. We will also explore the relationships between the variables, visualise trends, perform exploratory analysis to deliver the gained insights to a broader audience. I’ll be focusing particularly on the issue of racial bias by examining the dataset that may shed light on why the black people were represented as perpetrators of crimes.

Proportions of Incident Reasons by Subject’s Race

Starting off with the tables that depicts the proportion of incidents per subject’s race. The incident reasons include a variety of different reasons such as arrest, traffic stops, warrant execution, etc. The most number of events are occured for Black and Hispanic subjects. Particularly the common reasons for the incidents on Black subjects were arrest and service calls. The incidents on Hispanic subjects were mostly the same as black subjects. We also have other racial subjects with majorly arrests as the main reason for incidents.

We can interpret from the counts that the cases are most among the black subjects and the table also explains the relative frequency of incidents across different races.

SUBJECT_RACE INCIDENT_REASON Counts Percentage
American Ind Service Call 1 100.00%
Asian Arrest 4 80.00%
Asian Traffic Stop 1 20.00%
Black Arrest 673 50.49%
Black Call for Cover 71 5.33%
Black Crime in Progress 48 3.60%
Black Crowd Control 4 0.30%
Black NULL 6 0.45%
Black Off-Duty Employment 19 1.43%
Black Other ( In Narrative) 43 3.23%
Black Pedestrian Stop 26 1.95%
Black Service Call 360 27.01%
Black Suspicious Activity 28 2.10%
Black Traffic Stop 48 3.60%
Black Warrant Execution 7 0.53%
Hispanic Arrest 244 46.56%
Hispanic Call for Cover 32 6.11%
Hispanic Crime in Progress 23 4.39%
Hispanic Crowd Control 1 0.19%
Hispanic NULL 5 0.95%
Hispanic Off-Duty Employment 17 3.24%
Hispanic Off-Duty Incident 6 1.15%
Hispanic Other ( In Narrative) 13 2.48%
Hispanic Pedestrian Stop 3 0.57%
Hispanic Service Call 140 26.72%
Hispanic Suspicious Activity 11 2.10%
Hispanic Traffic Stop 28 5.34%
Hispanic Warrant Execution 1 0.19%
Other Arrest 5 45.45%
Other Off-Duty Employment 1 9.09%
Other Off-Duty Incident 1 9.09%
Other Service Call 4 36.36%
White Accidental Discharge 1 0.21%
White Arrest 215 45.74%
White Call for Cover 24 5.11%
White Crime in Progress 9 1.91%
White Off-Duty Employment 13 2.77%
White Off-Duty Incident 4 0.85%
White Other ( In Narrative) 12 2.55%
White Pedestrian Stop 7 1.49%
White Service Call 161 34.26%
White Suspicious Activity 8 1.70%
White Traffic Stop 14 2.98%
White Warrant Execution 2 0.43%

Scatter plot of Racial Disparities with number of incidents

The following plot shows the relation between the officers’ race and the subjects’. Each point in the plot shows the combination of officer and subject race with the size representing the number of incidents, the labels are added for better justification of incident counts. As we can see that the most force used on black subjects was by the white officers. This can provide insights into the potential racial disparities in the use of force by law enforcement. Comparatively, the crimes committed by the black subjects are way more than all other races collectively. We can also observe that white police make up the bulk in the departments, thus its no wonder why the treatment of black subjects is the way it is.

Racial Distribution by Gender

The following Pie charts represents the percentage of male and female subjects of each races. In the first one, each slice of the chart shows the percentage of male subjects with the label representing the exact numbers and following by the female subjects in the second one.

The pies of the subject’s race says it all, that most of the crimes are by the black subjects, followed by the hispanic subjects in males. Even when considering the female subjects, the majority crimes are by the black subjects and also some of the white subjects. Proportionately, the amount of subjects doesn’t differ much in terms of the gender.

Now, presenting the charts that depicts the distibution of Officer race by gender. Each pie’s size in the chart is proportional to the count of officers of that race. The color of the charts explains the race of officers.

When it comes to the races of the officers in the department, the white people dominates the crowd. The number of white officers is higher than the other races in the entire population in both the gender. The American Indians are the least to be in the police department. The percentage of the black officers is more in famale compared to the male.

Crime Rates in Dallas

The histogram at the left shows the overall crime rates in Dallas divided by divisions. It can be observed that the highest number of crimes are from the central division followed by the Northeast and the Southeast regions. The least number of crimes was reported from the Northwest division. The histogram at the right shows the crimes committed by the black subjects. Maximum reports were from the Central followed by the South Central divisions. The lowest number of crimes committed by black subjects is from Northwest division.

We can interpret from the plot that most of the cases were from the Central division area of the county. As we comprehend further the number of black subjects crimes in the county, more than half the other subjects crimes were committed by the black subjects, which strongly supports our point earlier in this report.

Crime Patterns in Dallas over the year

The following box plots explore the patterns of different factors on crime occurrences in Dallas.

The first box plot shows the distribution of the counts of crime in Dallas based on the subject’s Race. The plot compares the crime occurrences with the subject’s race, which are the boxes, and the black subjects has the highest median comparing to all other races. The plot tells that the American Indians & the Asians are the least crime committing races. The White and Hispanic subjects are quite moderate in terms of their crime rate.

As the black subjects are dominating in Dallas with their crimes, I wanted to take a deeper look on their level of misdeeds, So presenting a plot of black subjects in specific and their crime levels over the year. This compares the monthly average incident amount, denoted by the boxes, with the months in the x-axis. The rates varies over the year and can also spot some outliers in the black subjects plot.

From the plot, we can infer that the crimes are on the rise at the beginning of the year which slowly depreciates over months and slightly inclines around may and September. The amount of offenses is significantly lower at the end of the year than it was at the beginning. Altogether these plots, provide insights to a better extent and will be much helpful for the law enforcement agencies to understand and identify the crime rates and deploy suitable resources at the right time to reduce the crime rates.

Correlation Analysis

This analysis explores the relationship between the Incident reason and the reason for force charged on the subjects. Calculating its correlation value, which is +0.25, we can say that there is a positive correlation between the two. From the value we interpret that they have a moderate correlation, indicating that the increase in Incident reason increased the reason of force.

To visually explore the relation, a scatterplot has also been plotted using the values. This gives the association between the Reason of force applied for the Incident Reason. Each point denotes a different scenario, with Incident reason in x-axis and the Reason of force in y-axis. The plot shows that the cluster of points in area, indicates certain reasons for force on the particular subject due to an underlying incident reason. Mostly the reason behind the force was that the subject resisting Arrest and the least time was due to an Accidental Discharge in the concern of danger to self or others.

## Correlation between incident reason & reason for force: 0.2588171

Time Series Analysis

The plot consists of three line graphs denoting the occurrences of crimes over different time period.

The first plot depicts the hourly crime rate in Dallas, with counts in y-axis and the hour of the day in x-axis. The frequency of crimes is highest in the late evenings and early morning till 6 am. Gradually the amount of crimes decreases over the day time and inclines after 3 pm.

The second plot shows the daily frequency of crimes with Day in x-axis and crime counts in y-axis. The funny thing is that the most frequent of crimes happens at the weekends and consistent on the weekdays. The number of cases begin to incline on Wednesdays and improves over the weekend and abruptly decreases to the lowest level on Mondays.

The final plot is the amount of crimes over the year marked every month. As the plot represents the crime rates opens around 220 in January, slowly increases till March then decreases a bit gradually till July and achieves an amount of 200, then entirely goes down to 100 or less than that at the end of the year. The frequencies are at its peak in winter months.

This graph also represents the crimes over the year but marks every single day in a month, so we can gain insights even more deeper, with months in x-axis and crime counts in y-axis. As the lines are so irregular, smoothing them using geom_smooth, which helps to interpret it easily. The number of crimes peaks in a day of October 2016. This suggests that there may be some seasonal variation in the occurrences of crimes. We can’t be sure that this pattern repeats every year, because we only possess an year on data.

Crime zones in Dallas

This graph shows the map of Dallas and the corresponding crimes zones marked in the county using the latitudes and longitudes. The crime data has been filtered to focus on the Black subjects committing crimes around the city. As you can see the orange dots which represents the crimes committed by black subjects and the blue dots depicting the remaining subjects. The size of the points denotes the number of incidents at that particular location. The amount of crimes committed by the black subjects is higher when compared to the crimes collectively committed by other subjects.

Officers’ injury and their Experience in years

The plot shows the relation between the officer’s year of experience in the department and their amount of injuries. The x-axis represents the number of years of experience the officers have and the y-axis depicts the number of injuries the officers suffered over the year during their career. The graph illustrates that officers with less years of experience have higher instance of injury and the vice versa with the experienced officers. This could be due to multiple factors like, the candidates with fewer experience are more like vulnerable and mostly have less exposure on the crime scenarios.

Majority Type of Force1 used on Subjects

The plot shows the majority type of force1 used on criminals, which is divided by race and gender, respectively, to examine any force applied based on prejudice. As we can see that for both male and female subjects, the majority of type1 used force is “Verbal Command”, followed by the “Weapon display at person”. The least extreme action taken only on a black subject was “OC Spray”. From the graph we can clearly see that the actions were mostly taken on black subjects, which means that they can be so dangerous. Further investigation describes that the male subjects are most likely to experience “Take Down - Arm”, which is absent among female subjects.

The plot shows the majority type of force1 used by Officers, which is divided by race and gender, respectively, to understand which force will be applied for the first time by each officer in terms of their races and gender. Seemingly, they begin by “Verbal command”, that’s why its common with all the officers. A rare action which was only carried out by the White Officers was “Feet/Leg/Knee Strike”, which they applied on criminals who might have refused to surrender. In terms of the counts, we can confirm that most of the forces are applied by the white officers.

Subject Description per race

This interactive plot has been deployed using plotly, which shows the description of subjects along with their race. The plot displays the condition of the subject when the force was applied, along with their corresponding races. The size and color of the markers represents the number of subjects falling under each description. From the plot we can observe that the most cases were among the black subjects with majority of the description “Mentally Unstable”, followed by the “Unknown” status.

Conclusion

Based on the analysis of this policing data set with the use of force in Dallas, Texas, its evident that there are disparities in the way the police treat the individuals based on their race and gender. Also we can confirm through the plots we presented, that, the Black individuals were subjected to the use of force at a higher rate compared to white individuals, which makes them as perpetrators of crimes. Moreover the Male subjects were involved in the use of force than Female subjects. The data also showed that the police with more experience, had never got injured than the one with lesser experience.

In conclusion, the analysis of this data suggests that there is a need for further research to address disparities in use of force practices to individuals based on their gender and race. Also we should take some measures to reduce the crime rates by deploying more resources on suitable places and time periods.

References

Center for Policing Equity (2020). Data Science for Good: Center for Policing Equity - Police Data Initiative (Version 1.0) [Data set]. Kaggle. https://www.kaggle.com/center-for-policing-equity/data-science-for-good